Reflection 5 -- The use of Data Mining tools

In Lecture 4, we understand that Data Mining plays an important role in shaping better Customer Relationship Management (CRM) analysis. Accordingly, businesses should consider the use of Data Mining to enhance their CRM capabilities in order to be competitive globally.


However, in Lecture 5, Dr. Jagielska further assessed Data Mining in a more technical sense, where she used both Neural Networks as well as Decision Tree as the tools. In my opinion, there are 2 general issues related to the use of Data Mining tools, which will be evaluated below:


1. Privacy Issue
Privacy Issue may emerged as most customers are reluctant to give their personal information to the organization unless there are benefits for them in doing so. As a result, they may not participate or at most may even give misleading information that would directly affect the level of accuracy of the analysis.

2.Knowledge Issue
Within an organization, you can not expect an employee to be able to instantly understand how Data Mining tools operate due to the complexity of the software usage as well as the analysis that would be performed. As an example: To have high accuracy of analysis using a neural network software, an employee must have prior knowledge on how the Neural Network should be trained (This require thorough understanding on the arrangement of data, numbers of layers, and many more), what the input and output should be, and etc. To achieve best analysis with Neural Network software, companies must then spent a considerable amount of time and money in order to train the employee in using the software which would in turn enable organizations to make good decisions.

Now the main question would be: "Is it worth the cost?"

Both readings entitled "Rules are much more than decision tree" as well as "an overview of the CART Methodology" provided us with the insight on how both Neural Network and Decision Tree can be applied in the real world examples.

In my opinion, finding pattern of Data and/or predicting information using Decision Tree would be much simpler, easily understood and implemented as opposed to the Neural Networks. However, it is important to note that the Decision Tree would perform well when it is presented with little data. Accordingly, if the Decision Tree is given too much information, the level of the tree will be unmanageable where it would reduce the capability of Decision Maker to digest information in order to make good decisions.


Unlike Decision Tree, Neural Networks can be more complex to be implemented. However, when it is compared to the Decision Tree, the Neural Network would be more suitable for complex Data analysis due to its ability to process/analyze large amount of Data (Imagine creating Decision Tree with hundreds or thousands of Data). With Neural Network, a considerable amount of time and money need to be spent in training as it would require the analyst to have background understanding of Neural Network, how to best design its algorithm, as well as how to train and maintain them in order to have high accuracy when they are presented with sets of data that require analysis.

Both tools (Neural Networks and Decision Tree) have been adopted by several Large Organizations such as: Telephone Industry, Banking & Finance, as well as for Medical purposes. Looking at their future, further enhancement needs to be done in order to enable them to work efficiently and accurately when presented with events (such as: credit risk assessment, Health Assessment, and etc).

Feel free to comment on this post :)

Cheers, Yovita